MOLTO Project to Develop Multilingual Translation Tool
With a €2.3m fund, the European Union assists the development of a multilingual translation tool together with three renowned universities and industry partners across Europe. The three-year project is scheduled to be completed and be made available to the public in the form of open source libraries during 2013. A majority of the EU's official languages will be used for building the protoypes for MOLTO (Multilingual On-Line Translation).
Researchers from European Countries are working under the coordination of Aarne Ranta from the Department of Computer Science and Engineering at the University of Gothenburg, Sweden.
In a press statement, the MOLTO project coordinator said that it was the plan to allow users on completion of the project to freely download the tool and translate texts into several languages simultaneously. He pointed out that the technology to do this existed already, however he felt that it was cumbersome to use "unless you are a computer scientist".
It will be interesting to see how MOLTO will change how Europeans use machine translation, and whether the tool is as easy to use as envisaged. If existing translation assets from the European Union are used to their full capability it surely will add quality to machine translation for particular domains and subjects. Legislation and information for EU citizens are currently available in 23 languages and have to be updated and amended regularly, which results in enormous text corpora.
The FP7 funded project aims at developing a multilingual translation tool to support high quality machine translation of web content in real time. MOLTO uses statistical elements as its main technique such as domain-specific semantic grammars and ontology-based interlinguas. Research will be focused on the extension of rule-based translation by statistical methods.
Traditionally, machine translation tools were created in a rule-based structure. Put in a very simple context this means that as a rule, one word would be substituted with its equivalent translation of the very word into another language. Rules have been extended over the years to allow a recreation of linguistic rules such as understanding collocations and basic sentence structures.
The newer generation of machine translation includes statistical data, such as "which of the three possible translations for 'word' is most likely to occur next to word XZY in the context of ABC". This data is generally selected by the use of huge text corpora. A popular machine translation tool using this hybrid functionality of rules, statistics and to some extend content is Google's Translation Tool. The corpora used consists of material made available by the UN and other sources.The question is whether downloading, correctly installing an open-source library, and then plugging this library into other translation tools and websites will be easier for the users of machine translation than using the 'Translate' button in their browser's toolbar.
It will be a while until MOLTO is available, and undoubtedly machine translation will be used in the meantime by many individiduals and companies. Easy and inexpensive as it is, it has to be taken into account that is after all a translation by a machine and as such should always be professionally proofread if used for publication or communication purposes.
Read more about the MOLTO Project
Research Position for PhD Students
